The predictive value of pretherapy [68Ga]Ga-DOTA-TATE PET and biomarkers in [177Lu]Lu-PRRT tumor dosimetry

Abstract

Purpose: Metastatic neuroendocrine tumors (NETs) overexpressing type 2 somatostatin receptors are the target for peptide receptor radionuclide therapy (PRRT) through the theragnostic pair of ⁶⁸Ga/¹⁷⁷Lu-DOTATATE. The main purpose of this study was to develop machine learning models to predict therapeutic tumor dose using pre therapy ⁶⁸Ga -PET and clinicopathological biomarkers.

Methods: We retrospectively analyzed 90 segmented metastatic NETs from 25 patients (M14/F11, age 63.7 ± 9.5, range 38-76) treated by ¹⁷⁷Lu-DOTATATE at our institute. Patients underwent both pretherapy [⁶⁸Ga]Ga-DOTA-TATE PET/CT and four timepoints SPECT/CT at ~ 4, 24, 96, and 168 h post-¹⁷⁷Lu-DOTATATE infusion. Tumors were segmented by a radiologist on baseline CT or MRI and transferred to co-registered PET/CT and SPECT/CT, and normal organs were segmented by deep learning-based method on CT of the PET and SPECT. The SUV metrics and tumor-to-normal tissue SUV ratios (SUV_TNRs) were calculated from ⁶⁸Ga -PET at the contour-level. Posttherapy dosimetry was performed based on the co-registration of SPECT/CTs to generate time-integrated-activity, followed by an in-house Monte Carlo-based absorbed dose estimation. The correlation between delivered ¹⁷⁷Lu Tumor absorbed dose and PET-derived metrics along with baseline clinicopathological biomarkers (such as Creatinine, Chromogranin A and prior therapies) were evaluated. Multiple interpretable machine-learning algorithms were developed to predict tumor dose using these pretherapy information. Model performance on a nested tenfold cross-validation was evaluated in terms of coefficient of determination (R²), mean-absolute-error (MAE), and mean-relative-absolute-error (MRAE).

Results: SUV_mean showed a significant correlation (q-value < 0.05) with absorbed dose (Spearman ρ = 0.64), followed by TLSUV_mean (SUV_mean of total-lesion-burden) and SUV_peak (ρ = 0.45 and 0.41, respectively). The predictive value of PET-SUV_mean in estimation of posttherapy absorbed dose was stronger compared to PET-SUV_peak, and SUV_TNRs in terms of univariate analysis (R² = 0.28 vs. R² ≤ 0.12). An optimal trivariate random forest model composed of SUV_mean, TLSUV_mean, and total liver SUV_mean (normal and tumoral liver) provided the best performance in tumor dose prediction with R² = 0.64, MAE = 0.73 Gy/GBq, and MRAE = 0.2.

Conclusion: Our preliminary results demonstrate the feasibility of using baseline PET images for prediction of absorbed dose prior to ¹⁷⁷Lu-PRRT. Machine learning models combining multiple PET-based metrics performed better than using a single SUV value and using other investigated clinicopathological biomarkers. Developing such quantitative models forms the groundwork for the role of ⁶⁸Ga -PET not only for the implementation of personalized treatment planning but also for patient stratification in the era of precision medicine.

Keywords: 177Lu-DOTATATE; Dosimetry; Machine learning; Theragnostic.

Fig. 1

Top panel: baseline diagnostic images (contrast-enhanced CT/MRI) were used to define target lesions, which were then co-registered to pretherapy ⁶⁸Ga-DOTATATE PET/CT and posttherapy 177Lu-DOTATATE SPECT/CT images. Bottom panel: dosimetry pipeline included four timepoints registration of SPECT images to generate TIA that is fed into MC-based dose engine

Fig. 2

Spearman rank self and cross correlation between absorbed dose-related parameters (dose, TIA_vol, C_vol and T_eff) and PET-SUV parameters along with biomarkers. The color code and size of spheres show the correlation magnitude. The insignificant correlations ($q$-value > 0.05) are plotted as faded spheres

Fig. 3

Intra-patient variability of tumor absorbed doses per unit administered activity for all patients. The sphere color indicates SUV_mean, and background color shows the margins of standard deviation of tumor absorbed dose per unit administered activity values. The size of spheres depicts the volume of tumors in logarithmic form (4–1039 mL)

Fig. 4

Tumor absorbed dose plotted vs. tumor PET-SUV quantities, where the color shows the tumor location. The size of spheres depicts the volume of tumors in logarithmic form (4–1039 mL)

Fig. 5

Pretherapy predicted absorbed dose using univariate linear model and random forest (RF) bi/tri-variate models of Table 2 vs. the delivered dose measured from Lu-177 SPECT/CT (the filled gray dots represent the 2 outliers)

Fig. 6

Considering the threshold absorbed dose for responders of 25 Gy/cycle, confusion matrix of predicted absorbed dose from trivariate Ens-Tree model compared to the measured absorbed dose (left). Sensitivity $\left(\frac{TP}{TP+FN}\right)$ and specificity $\left(\frac{TN}{FP+TN}\right)$ visualization of the predcition model (right). TP true-positive, FP false-positive, TN true-negative, FN false-negative